Taleb made many local mistakes in his books and went back and corrected almost all of them without anybody knowing.

This claim is about Taleb's own books and his own actions, so it is judged as of 2017-07-10. True: there is documented evidence that Taleb's books (the Incerto volumes) contained many small, local errors that were fixed in later printings or editions, that almost all of the known errors were fixed, and that the fixes were made quietly, with no published errata or announcement. Mostly true: there were many quiet fixes, but a minor detail is off, for example some fixes were briefly noted publicly or a small share of known errors stayed. Mixed: the errors were real and some were fixed, but a large share of documented errors stayed in later printings, or the fixes were openly announced (published errata lists), so either 'almost all' or 'without anybody knowing' clearly fails. Mostly false: few of the known errors were fixed, or the fixes were widely publicised. False: comparing editions shows no fixes to documented errors, or the books are shown to have had no such errors. If edition comparisons or publisher records cannot be found, the claim is unresolvable, not false.

Set before the research began.

1. E1 Fooled by Randomness - Wikipedia Wikipedia

   “In 2005, Random House published a softback edition with more changes.”

2. E2 Still fooled by randomness, twenty years later? Roskilde University research database (peer-reviewed book review by Spaniol, Evans, Tranekær, Rowland and Atkinson) · June 2021

   “Taleb's (2005) updated softback second edition adds 146 pages to the original hardback, and con- tains, we agree with Stone (2005), “writing that's a bit smoother.””

3. E3 Nassim Nicholas Taleb: Fooled by Randomness Grumpy Old Bookman (blog) · 20 August 2004

   “a revised and expanded second edition was issued earlier this year.”

E1, E2 and E4 show only that Fooled by Randomness came out in a revised and expanded second edition, with a 2005 softback that added 146 pages and had smoother writing. None of them names specific local errors in the Incerto books, says whether later printings fixed them, or says whether any fixes were made quietly or announced in errata. The criteria require edition comparisons or publisher records, and this evidence does not supply them, so the claim cannot be settled.